A microscopic tool being developed by this team will allow quantitative absorption and fluorescence 3D imaging of cells and their organelles which can be correlated to the clinical standards of disease diagnosis for the first time. This project develops new fields of 3D quantitative microscopy in two important contrast modes that bridge a gap between clinical diagnosis and medical research. In addition, the technology of 3D image processing from quantitative CT imaging for disease diagnosis is being applied to microscopy in a unique collaboration.
Textbooks of cell biology depict a cell as flat with two-dimensional features. If disease diagnosis is 3D and more quantitative and sensitive to disease onset, then these textbooks will likely change to more accurately represent the cell anatomy. Teaching students on cellular structures and functions may be significantly changed due to advancements of true 3D microscopy. Providing cytologists and molecular biologists with quantification tools will result in earlier diagnosis of lethal diseases like lung cancer and improved understanding of the disease mechanisms.
The three-dimensional (3D) Optical Projection Tomography Microscope (OPTM) is a new and exciting instrument for imaging small translucent objects such as single cells. This novel microscope facilitates the acquisition of full 3D optical density images of single biological cells. For medical imaging the advent of X-ray computerized tomography (CT) imaging allowed the transition from blurred two-dimensional chest radiographs to precise calibrated 3D images with high accuracy in both geometry and x-ray density. In a similar way and with a potential similar outcome the OPTM is an advance on two-dimensional (2D) conventional optical microscopy. In this case, the optical rather than X-ray density of the translucent 3D object is obtained. However, the method of reconstructing a 3D precise image from a set of 2D projections is similar in concept to the reconstruction of a 3D image in CT imaging from multiple chest X-ray images recorded from different viewing directions. The focus of this research project is to explore automated computer methods to analysis the OPTM images to assist in the diagnosis of lung caner. In cytopathology cancer diagnosis is made from images of single cells that have been obtained from the human body; for example, from blood, sputum, or biopsy. A number of cells are examined for each patient in order to find cells with cancer identifying abnormalities. This research builds on our previous experience in the analysis of 3D chest X-ray CT images to diagnose lung cancer by measurements on pulmonary nodules to discover robust computer methods for the new OPTM technology. The major research outcomes are: (a) a method to partition a cell image, (b) a method for quantitative volumetric measurement that is more precise than is possible with 2D microscopy, (c) a method to measure chromatin content comparable to flow cytometric methods, and (d) development of an initial cell classification system that can identify abnormal cancer cells from normal cells. There were two major research collaborations for this project. Eric Seibel’s research group at the University of Washington provided image data and collaborated on all phases of this project. VisionGate Inc. manufactures the OPTM instrument and provided cell images for outcomes (a) and (d) above. In summary, we have shown that the precise 3D information provided by the OPTM exceeds the capabilities of conventional 2D microscopy in several important aspects and, therefore, provides a potentially superior technology for lung cancer diagnosis. Critical to any cell analysis is the ability to identify the cell nucleus and cytoplasm components. We have developed completely automated computer analysis methods to decompose a 3D cell image into thee parts (a) the cell nucleus, (b) the cell cytoplasm and (c) the space outside the cell. From this partitioning we can make a number of measurements of cell nucleus and cytoplasm characteristics including: volume, density, and shape. Our new method is based on the convex hull and multiple applications of graph-cuts, which is more robust than traditional methods. We have shown that we can automatically measure the nuclear/cytoplasm ratio (i.e., the ratio of volume of the nucleus to the volume of the cytoplasm), which is used by cytologists as an indication of cell state. Further we have shown that this measure made on a 3D OCTM image is more precise than measurements made with traditional 2D microscope images. The DNA index of cancer cells was measured by the amount of light absorption within the cell nucleus. This measurement was made accurate with the use of internal calibration standards of bare cell nuclei. This new diagnostic measurement of cell aneuploidy is a strong biomarker for cancer. We have developed an automated cell classification system that identifies abnormal cancer cells from four other normal cell types. Initial results have been very promising this work is still being completed.
